On the choice of the optimal parameter for Tikhonov regularization of ill-posed problems (Q1209034)

Tikhonov's regularization method is used to solve the ill-posed equation \(Kx = g\), where \({K: H_ 1 \to H_ 2}\) is compact and \(\text{Range}(K)\neq\overline {\text{Range}(K)}\). The problem the author considers is the choice of Tiknonov's regularization parameter \(\alpha\), which is involved in the smoothing functional \(F_ 2(\alpha,z) = \| Kz-g^ \delta\|^ 2 + \alpha\| z\|^ 2\). The author proves that if \(p(q+1) = 2/(2+1)\), then \(\| x^ \delta_ \alpha - x\| = O(\delta^{2\nu/(2\nu H)})\) for \(x \in \text{Range}(K^*K)\), \(1/2 \leq \nu \leq 1\), and that if \(\| x^ \delta_ \alpha - x\| = O(\delta^{2\nu/(2\nu H)})\) for \(x \in \text{Range}(K^*K)\), \(0 < \nu \leq 1\), then \(p/(q+1) \in \left[{2\nu\over 2\nu+1},{2\over 2\nu+1}\right]\), where \(p\), \(q\) satisfy \(\| Kx^ \delta_ \alpha - g^ \delta\| = \delta^ p/\alpha^ q\) and \(x^ \delta_ \alpha\) is the minimizer of \(F_ 2(\alpha,z)\).